Loading ...
Sorry, an error occurred while loading the content.
 

Re: [Electronics_101] Toriods and Saturation

Expand Messages
  • Shawn Upton
    ... Ok, what about the primary then? I am still forcing exactly 2App into the primary, and the primary is still 50 turns. The flux density, load or no load,
    Message 1 of 26 , Oct 1, 2006
      --- John Popelish <jpopelish@...> wrote:

      >
      >
      > Nope, I don't think so. The voltage produced per
      > turn is proportional
      > to the rate of change of flux through the turn. If
      > the turn is short
      > circuited, and needs only the voltage to over come
      > its resistance, you
      > can produce large current with little voltage, and
      > thus, little rate
      > of change of flux. Integrate little rate of change
      > over a sine half
      > cycle and you get little total flux.

      Ok, what about the primary then? I am still forcing
      exactly 2App into the primary, and the primary is
      still 50 turns. The flux density, load or no load,
      that is generated is still uNI/l. Assuming a good
      transformer, NI is the same on either end.

      > I think you are worrying about something that is
      > insignificant in
      > parallel with a 2500 to 1 step up of the impedance
      > of a short circuit.

      I'll recheck tomorrow, by shorting the secondary and
      then rechecking the inductance of the primary at
      100kHz. I'm baffled by the operation of the driver
      though: I was able to get 1App at 100kHz to drive a
      very small inductive load (basically an 8:1 step-up
      but with an air gap, so relatively low inductance)
      albeit with a definate reduction in drive due to the
      driver running out of bandwidth. But I could not
      drive the 50:1 transformer worth beans past 20kHz; the
      voltage *and* current as seen into and across the
      primary simply dropped off. Not only that, but I had
      a significant amount of voltage required to get the
      prescribed 2App below this point--indicating high
      leakage inductance, not a good sign (low coupling).

      It may very well be this core is just lousy as the
      frequency is increased--the permability is simply
      dropping off too fast. I suspect that I simply need a
      new core, with a bit larger diameter (2 inch or more),
      similar cross sectional area (to keep the inductance
      reasonably high for low frequency operation), and a
      datasheet.

      > For instance, if the total impedance of the
      > secondary, the
      > connecting leads and the load is 10 milliohms, the
      > capacitance of the
      > primary is in parallel with 25 ohms. The capacitive
      > impedance of the
      > primary would have to be below 2500 ohms to bypass
      > at least 1% of the
      > source current around the transformer. That would
      > be about 640 pF at
      > 100kHz.

      Good point. I'll write down the value I find this
      time, as the LRC meter indicated some high value of
      capacitance at 100kHz when I measured (negative
      inductance, so I changed to capacitance and I want to
      say 1nF--something way in excess of what I'd expect).
      I'm not sure where the SRF is, but it's below 100kHz,
      which is a definate problem for me.

      Fun problem, more than I expected when I started
      digging into this. :)


      Shawn Upton, KB1CKT

      __________________________________________________
      Do You Yahoo!?
      Tired of spam? Yahoo! Mail has the best spam protection around
      http://mail.yahoo.com
    • John Popelish
      ... (snip) ... If it is a good transformer, Nprimary*Iprimary equals Nsecondary*Isecondary, except that these two are in opposite directions. uNI/l applies
      Message 2 of 26 , Oct 1, 2006
        Shawn Upton wrote:
        >
        > --- John Popelish <jpopelish@...> wrote:
        (snip)
        >>The voltage produced per turn is proportional
        >>to the rate of change of flux through the turn.
        >>If the turn is short circuited,
        >>and needs only the voltage to over come
        >>its resistance, you can produce large current
        >>with little voltage, and thus, little rate
        >>of change of flux. Integrate little rate of change
        >>over a sine half cycle and you get little total flux.
        >
        > Ok, what about the primary then? I am still forcing
        > exactly 2App into the primary, and the primary is
        > still 50 turns. The flux density, load or no load,
        > that is generated is still uNI/l. Assuming a good
        > transformer, NI is the same on either end.

        If it is a good transformer, Nprimary*Iprimary equals
        Nsecondary*Isecondary, except that these two are in opposite
        directions. uNI/l applies only if the secondary is open circuited,
        when the voltage drop of the primary is caused by its inductive
        impedance, not the transformed short from the secondary. In that
        case, the core would definitely see a lot of flux swing, especially if
        your current source could actually supply enough voltage to force the
        same current through it.

        >
        >>I think you are worrying about something that is
        >>insignificant in
        >>parallel with a 2500 to 1 step up of the impedance
        >>of a short circuit.
        >
        >
        > I'll recheck tomorrow, by shorting the secondary and
        > then rechecking the inductance of the primary at
        > 100kHz. I'm baffled by the operation of the driver
        > though: I was able to get 1App at 100kHz to drive a
        > very small inductive load (basically an 8:1 step-up
        > but with an air gap, so relatively low inductance)
        > albeit with a definate reduction in drive due to the
        > driver running out of bandwidth. But I could not
        > drive the 50:1 transformer worth beans past 20kHz; the
        > voltage *and* current as seen into and across the
        > primary simply dropped off.

        Substitute a big 25 ohm resistor for the primary and see how your
        source responds to that.

        > Not only that, but I had
        > a significant amount of voltage required to get the
        > prescribed 2App below this point--indicating high
        > leakage inductance, not a good sign (low coupling).

        An air gapped transformer core has a huge leakage inductance between
        primary and secondary, compared to an ungapped, high permeability core.

        > It may very well be this core is just lousy as the
        > frequency is increased--the permability is simply
        > dropping off too fast. I suspect that I simply need a
        > new core, with a bit larger diameter (2 inch or more),
        > similar cross sectional area (to keep the inductance
        > reasonably high for low frequency operation), and a
        > datasheet.

        The lowest cost, readily available cores with high permeability to
        over 100 kHz I know of are the LFB series of Steward cores from Digikey.

        >> For instance, if the total impedance of the
        >>secondary, the
        >>connecting leads and the load is 10 milliohms, the
        >>capacitance of the
        >>primary is in parallel with 25 ohms. The capacitive
        >>impedance of the
        >>primary would have to be below 2500 ohms to bypass
        >>at least 1% of the
        >>source current around the transformer. That would
        >>be about 640 pF at
        >>100kHz.
        >
        >
        > Good point. I'll write down the value I find this
        > time, as the LRC meter indicated some high value of
        > capacitance at 100kHz when I measured (negative
        > inductance, so I changed to capacitance and I want to
        > say 1nF--something way in excess of what I'd expect).
        > I'm not sure where the SRF is, but it's below 100kHz,
        > which is a definate problem for me.
        >
        > Fun problem, more than I expected when I started
        > digging into this. :)

        Yes, lots of fun.
      • Shawn Upton
        Ah, took me a moment--for powdered iron cores, the higher the permability, the more conductive the core is. The core I m playing with is ferrite though, and I
        Message 3 of 26 , Oct 7, 2006
          Ah, took me a moment--for powdered iron cores, the
          higher the permability, the more conductive the core
          is. The core I'm playing with is ferrite though, and
          I don't think they are conductive as a rule. Haven't
          read up on their construction yet though. My core is
          like mu r = 10,000, based upon some measurements.

          I experimented around, and indeed I have not been able
          to saturate my core, even at stupid high currents.
          There is something that I need to go back and look at,
          a pecularity that I couldn't quite figure out.
          Basically, I figure, if the core saturates, I should
          see flattopping in the current: flux doesn't increase
          as fast as it should in saturation (core saturation is
          sudden not gradual), and so current transfer on the
          peaks should clip in a soft sort of way. Instead, it
          looked like zero crossing error in the current. Could
          be the driver.

          Next week, I think I'll unhook the secondary and then
          watch V vs I and see if I can detect a saturation
          level, and then rehook the secondary and see if it
          saturates at the same current. My understanding is
          obviously wrong about this transformer, in terms of
          saturation.

          A fun side note: I attempted to make my own resistor
          to measure current, using 23mm of 18g wire. 0.5
          milliohm. I was able to get what I think was around
          270App at 1kHz, not bad--but the shunt managed to
          unsolder the scope probe at that power level!

          [I had used RG-174 as a probe to the scope, that way I
          could measure mV's. Works well. But the coax braid
          went to the 3/8" braid I was using as a secondary, and
          the shunt was connected to the DUT, and thus the
          center conductor could get hot enough to unsolder
          itself. It's a fun thing, to be able to smell your
          circuit, and know it's working. Maybe I should go
          work with some tubes in the future.]

          Anyhow, I ran a bunch of tests, and then went back to
          figure out if I could "trust" my shunt. Well, I
          forget what inductance I calculated for 23mm of wire,
          but in the end, at 5kHz, I found the wire to have 0.5
          milliohm of inductive reactance. Not so good, means
          the current waveform as measured isn't what I thought!
          Not only that, but above 10kHz, the skin effect
          starts to crop up too, increasing the resistance.

          In the end, I'm shifting gears. I'm going to make a
          push pull pair, using IRF510's, to drive the
          transformer with a different set of windings (probably
          6 or 7:1). OPA627's to drive the FET's, one as an
          integrator and the other as an invertor, so as to
          drive the gates out of phase. 0.1 ohm shunt, so as to
          stay resistive past 100kHz, and some feedback so that
          the response is flat.

          It simulated well. Looked good, was about to build it
          until I realized that the DUT is inductive. Adding in
          the 60nH of inductance causes a nice oscillation, so I
          haven't been able to build it yet. The slight phase
          shift between the 60nH inductor and the large phase
          shift from the gate capacitance (reacting with the
          usual gate resistor) causes a 1.2MHz oscillation.
          Drat, was so close! Haven't been able to get a lead
          network to tame it yet either.

          Shawn

          --- John Popelish <jpopelish@...> wrote:

          > --- In Electronics_101@yahoogroups.com, Shawn Upton

          > If you have several candidate cores on hand, you can
          > easily find the
          > high permeability ones that will provide highest
          > winding coupling by
          > touching them on two spots with your ohm meter
          > leads. The lower the
          > resistance reading, the higher the permeability, as
          > a general rule.
          >

          Shawn Upton, KB1CKT

          __________________________________________________
          Do You Yahoo!?
          Tired of spam? Yahoo! Mail has the best spam protection around
          http://mail.yahoo.com
        • John Popelish
          ... Powdered metal cores are almost all good insulators, because the bonding material between the grains is good insulation. Only a few very high permeability,
          Message 4 of 26 , Oct 7, 2006
            Shawn Upton wrote:
            > Ah, took me a moment--for powdered iron cores, the
            > higher the permability, the more conductive the core
            > is. The core I'm playing with is ferrite though, and
            > I don't think they are conductive as a rule. Haven't
            > read up on their construction yet though. My core is
            > like mu r = 10,000, based upon some measurements.

            Powdered metal cores are almost all good insulators, because
            the bonding material between the grains is good insulation.
            Only a few very high permeability, low frequency materials
            have significant conductivity. They are also easily
            recognized by colored paint and a gray, grainy appearance if
            chipped. Ferrite looks black and glassy if chipped.

            There are two basic families of ferrite in wide use. The
            high permeability family, including the high flux power
            supply core types are made of manganese ferrite and are semi
            conductive. The high frequency types, including most f the
            EMI suppression cores (except for a few exceptions) are made
            of nickel zinc ferrite and are pretty good insulators.
            They are such good insulators, that you can't sort them with
            an ohm meter, except to tell that they are nickel zinc. But
            the manganese zinc family have the general property that the
            higher the permeability (and lower the frequency capability)
            the lower the resistivity, and can be fairly accurately
            sorted with an ohm meter. The data sheets for the various
            materials should include a spec for the bulk resistivity.

            > I experimented around, and indeed I have not been able
            > to saturate my core, even at stupid high currents.
            > There is something that I need to go back and look at,
            > a pecularity that I couldn't quite figure out.
            > Basically, I figure, if the core saturates, I should
            > see flattopping in the current:

            I thought you were driving your transformer with a current
            source. If so, the primary current should be independent of
            the cores state. The output current will diminish as the
            core saturates, because more and more of the primary current
            will be going into stored magnetic energy as the core saturates.

            (snip account of surprises)
            > A fun side note: I attempted to make my own resistor
            > to measure current, using 23mm of 18g wire. 0.5
            > milliohm. I was able to get what I think was around
            > 270App at 1kHz, not bad--but the shunt managed to
            > unsolder the scope probe at that power level!

            Sounds like you need 4 pieces, 4 times as long, in parallel.
            Fold each into a hair pin to get the ends close together,
            and keep the inductance down.

            (snip hot ware story)
            > Anyhow, I ran a bunch of tests, and then went back to
            > figure out if I could "trust" my shunt. Well, I
            > forget what inductance I calculated for 23mm of wire,
            > but in the end, at 5kHz, I found the wire to have 0.5
            > milliohm of inductive reactance. Not so good, means
            > the current waveform as measured isn't what I thought!
            > Not only that, but above 10kHz, the skin effect
            > starts to crop up too, increasing the resistance.

            I would be thinking of something like a square of brass shim
            stock with a heavy copper bar soldered across each end as
            connection points. You can punch the middle with a paper
            punch to get close and then trim slivers off the sides.

            > In the end, I'm shifting gears. I'm going to make a
            > push pull pair, using IRF510's, to drive the
            > transformer with a different set of windings (probably
            > 6 or 7:1). OPA627's to drive the FET's, one as an
            > integrator and the other as an invertor, so as to
            > drive the gates out of phase. 0.1 ohm shunt, so as to
            > stay resistive past 100kHz, and some feedback so that
            > the response is flat.
            >
            > It simulated well. Looked good, was about to build it
            > until I realized that the DUT is inductive. Adding in
            > the 60nH of inductance causes a nice oscillation, so I
            > haven't been able to build it yet. The slight phase
            > shift between the 60nH inductor and the large phase
            > shift from the gate capacitance (reacting with the
            > usual gate resistor) causes a 1.2MHz oscillation.
            > Drat, was so close! Haven't been able to get a lead
            > network to tame it yet either.

            Have you tried a fat ferrite bead over the source leads of
            the fets? A low value wire wound source resistor might also
            help.
          • Shawn Upton
            ... Yep, still using a current source. I wanted to remove the effect of the secondary, thus I can only watch the voltage across the inductor to see what
            Message 5 of 26 , Oct 7, 2006
              --- John Popelish <jpopelish@...> wrote:

              >
              > I thought you were driving your transformer with a
              > current
              > source. If so, the primary current should be
              > independent of
              > the cores state. The output current will diminish
              > as the
              > core saturates, because more and more of the primary
              > current
              > will be going into stored magnetic energy as the
              > core saturates.

              Yep, still using a current source. I wanted to remove
              the effect of the secondary, thus I can only watch the
              voltage across the inductor to see what happens. That
              way, I can see if the flux cancelation (due to
              transformer action) does make an impact.

              I'd think that the voltage would clip if the core
              material saturates--if the flux cannot build any
              higher (no more stored field), then the voltage should
              clamp or even droop, even as current goes higher.

              > Sounds like you need 4 pieces, 4 times as long, in
              > parallel.
              > Fold each into a hair pin to get the ends close
              > together,
              > and keep the inductance down.

              That's a good idea, hadn't thought of it. Thanks.

              >
              > Have you tried a fat ferrite bead over the source
              > leads of
              > the fets? A low value wire wound source resistor
              > might also
              > help.
              >

              No, haven't built it yet, just simulating first in
              LTSPICE, to check for problems. And problems show up,
              so I'm hesitant to build just yet.

              The FET's have a 20 ohm gate resistor, originally I
              put them in there just to prevent the op-amps from
              oscillation (with the gate capacitance). I tried to
              insert some inductance, but I couldn't get it stable,
              and the gate voltage starts swinging real hard. I've
              seen RF amps take more than the spec'd +/-20V gate
              limit, but I'm not so sure at these relatively low
              frequency levels.

              I also tried a lead network here on the gate, still no
              luck (RC across the gate resistor, to try to get some
              positive phase shift). Frequency of oscillation just
              moves around.

              I tried to add some lead to the feedback, but that
              hasn't helped either. I think I have to sit down and
              figure out the phase response over frequency, and do
              the math and see what I actually need to do in order
              to compensate this: I tried some "standard" fixes here
              and there (bigger gate resistor, caps here and there)
              to no success.

              Shawn Upton, KB1CKT

              __________________________________________________
              Do You Yahoo!?
              Tired of spam? Yahoo! Mail has the best spam protection around
              http://mail.yahoo.com
            Your message has been successfully submitted and would be delivered to recipients shortly.